Download Risk and Uncertainty in Macroeconomic Policymaking

Robert Kirkby and Martin Fukac
Managing and
Communicating
Risk and Uncertainty in
Macroeconomic Policymaking
Introduction
In policymaking, as in life, we often must make decisions
without knowing how the future will play out. Taking
uncertainty into account when making macroeconomic
policy allows policymakers to help improve economic
outcomes. This article considers three aspects of improving
outcomes for fiscal and monetary policymaking: considering
risk and uncertainty in decision-making; communicating risk
and uncertainty; and designing better tools to communicate
risk.
To start with, it is useful to differentiate
between risk and uncertainty, terms that
are often used interchangeably.1 Risks are
what we might call the ‘known unknowns’.
They are future events for which the
past provides guidance on both their
likelihood of occurring and their effects,
and we can insure ourselves against them.
An example of a macroeconomic risk for
the New Zealand economy is exchange
rate movements: there is a long history of
exchange rate movements which we can
use as a basis for assessing the likelihood
of small or large changes in the future.
Robert Kirkby is a Lecturer in the School of Economics and Finance in Victoria Business School at
Victoria University of Wellington. Martin Fukac is a Principal Advisor in the New Zealand Treasury.
Page 40 – Policy Quarterly – Volume 12, Issue 3 – August 2016
Risk lends itself to measurement and
quantification by statistical tools. Using
such tools is a useful and important part
of policymaking. Further, the results of
such measurement and quantification
can, and often should, be communicated
to the public.
By contrast, uncertainty captures
the ‘unknown unknowns’, often called
Knightian uncertainty following the
pioneering work of Frank Knight
(1921). Uncertainty is about events
that cannot be foreseen or defined a
priori. Their likelihood of occurrence
and macroeconomic impacts are
not quantifiable because the past is
considered to provide little guide to the
future. Hence, statistical tools cannot be
used to evaluate the likelihood of future
outcomes. By definition it is difficult to
discuss specific uncertain events, and
so we focus here on uncertainty as a
concept. If we can describe likely events
and the damage they may then do, they
are risks, not uncertainties.
But how should we react to the genuine
uncertainties? Optimal behaviour when
faced with uncertainty is to minimise the
potential damage from worst possible
outcomes. This is known as maximin (i.e.
maximize the minimum) in the technical
literature. But how should this concept be
applied in practice? A good response is to
think about strategies to build resilience
capable of cushioning against adverse
outcomes. We may not know what all the
uncertainties are, but we can think about
our ability to absorb these uncertainties
as they unfold. Some uncertainties will
be positive, and these can be treated as
windfalls. In this vein, we will describe
how such resilience might be evaluated
and communicated as a worthwhile
answer to the difficulties of describing
uncertainties and the impossibility of
quantifying them.
Dealing with risk and uncertainty
is a fruitful area for advancing better
government
through
collaboration
between academics and those who make
policy. It requires an understanding of the
available tools with their strengths and
weaknesses, an area in which academia
has a definite strength, while government
advisers and decision-makers bring an
understanding of where such analyses
might be applied to inform and improve
policy.
How do policymakers take risk and
uncertainty into account when making
decisions?
Macroeconomic policies are often made
with the goal of reducing economic
fluctuations and the risks around them.
Uncertainties around future policy
reactions can be converted into risks by
outlining how policymakers would react
in various scenarios (Ilut and Schneider,
2014), to reduce risk around the policies
themselves (Fernández-Villaverde et al.,
2015)2 and to reduce the size of economic
fluctuations (Bekaert, Hoerova and Duca,
2013). Lowering economic and policy risk
by identifying possible risks and using
them to inform policymaking cultivates
an economic environment in which
businesses tend to invest more and employ
more people (Bloom, Bond and Van
Reenen, 2007; Bloom, 2009). Likewise, if
consumers perceive income certainty they
will tend to spend more (Bertola, Guiso
and Pistaferri, 2005). Conversely, more
risk – other things being equal – means
less income, less investment and more
unemployment. Reducing risk can thus
help to both increase the level of GDP and
reduce fluctuations in GDP.3
In New Zealand, as one of the two
pillars of macroeconomic policy, fiscal
policy contributes to reducing economic
uncertainty by keeping public finances
in order, sustaining a stable tax system
and ensuring predictable expenditure
policies. As the second pillar, monetary
policy contributes to reducing economic
uncertainty by maintaining price stability
through a transparent and predictable
interest rate policy.
In practice, policymaking by monetary
and
fiscal
authorities
for
macroeconomic
risk
management
purposes typically consists of four basic
steps: the identification and quantification
that policy decisions refer to an assessment
of risks and alternative scenarios that
have been considered. On the other hand,
we seldom find in available documents
formal probabilistic assessment of
projected economic conditions or cost–
benefit analysis. We are among those who
argue that more can be done in these areas
in New Zealand. We return to this point
in the final sections of this article, where
we talk about tools for communicating
risk outlooks and cooperation between
government institutions and academia.
We often look to the past for guidance
on which policies appear to work or not
work in order to make better policy.
Whether this is done informally through
discussion or formally as a statistical
analysis of past policies, it is important
that we are aware that the past was, at the
time, a risky future. Orphanides (2001)
shows how the uncertainty in the realtime data has real consequences if we
... we seldom find in available
documents formal probabilistic
assessment of projected economic
conditions or cost–benefit analysis.
of risk; decisions about whether to
mitigate risks; decisions about whether to
make provision for risks; and decisions
about whether to accommodate for
residual risks (IMF, 2016). A variety of
analytical tools support those decisionmaking processes. A typical set of
heuristic methods includes: general
probabilistic assessment of current and
projected economic conditions based on
historical risk valuation; the classification
and valuation of risk that can be
controlled or reacted to; assessments of
specific alternative risk scenarios, and
which risks are simply beyond the control
of policymakers; and the cost–benefit
analysis of policy options.
In international comparisons, many
observers rank New Zealand practices
among the most advanced (for example,
on the monetary policy framework see
Svensson, 2009; for fiscal policy see TerMinassian, 2014). The standard practice is
ignore it and simply look at historically
revised data when evaluating monetary
policy decisions of the past. For example,
with hindsight, revised data may show
that a recession was not as large as was
thought at the time and that monetary
policy was overly loose.
Risk analysis can be very extensive,
and for those interested in an academic
reading on policy and uncertainty we
suggest Public Policy in an Uncertain
World: analysis and decisions by Charles
Manski (2013). However, there is a limit
to what policy can achieve with respect
to reducing and managing risks. Kydland
and Prescott (1977) teach us that good
policy does not need to, and often
should not, try to react to every single
development.
On the other hand, no matter how
many risk scenarios policymakers
consider, they will still face uncertainty
– the ‘unknown unknown’. What should
Policy Quarterly – Volume 12, Issue 3 – August 2016 – Page 41
Managing and Communicating Risk and Uncertainty in Macroeconomic Policymaking
policymakers do when faced with such
uncertainties? There are two angles to
the answer. The first is operational:
what is the first best response when the
unknown unknown materialises? To
minimise mistakes in such situations,
robust control theory recommends that
policymakers follow heuristics (‘rules
of thumb’). Rules of thumb support
robust decisions that help minimise the
ex post adverse outcomes in the case of
uncertainties (see, for example, Dupuis,
James and Peterson, 2000; Hansen and
Sargent, 2001).
The second response is building the
resilience of the economy. Instead of only
trying to consider what uncertainties
the economy faces, we might also
consider how to improve the ability of
they are unknown and unknowable.
However, we can discuss the concepts of
uncertainty and resilience. Policymakers
can discuss resilience and weaknesses
(we will later describe ways to measure
resilience). Thinking about how to
deepen resilience and remove weaknesses
provides a sensible strategy for dealing
with uncertainty.
Should risk and uncertainty be
communicated?
The potential power of communicating
policy is captured in the term ‘open mouth
operations’ coined by former Reserve Bank
of New Zealand governor Don Brash.
He observed that he seemed to be able
to move interest rates simply by talking,
without conducting the actual open
[A case for] communicating risks is
simply that the analysis of these risks will
often already have been undertaken as
part of policymaking, and this information
may be valuable in and of itself.
the economy to ‘roll with the punches’.
An example in fiscal policy is the idea
of ‘fiscal space’: that is, ensuring that the
country is positioned such that should
adverse uncertainties become manifest, it
is possible to implement fiscal policies to
counter them. The experience of Ireland
and Spain during the recent global
financial crisis illustrates the value of
such ‘fiscal space’: both countries found
themselves in a position where fiscal
austerity was necessary, despite being
undesirable during a recession; if they
had gone into the crisis with greater fiscal
space and resilience they would have been
able to avoid this. The aim of such policies
should be to enhance the resilience and
adaptability of the economy to absorb
adverse economic shocks arising from the
‘unknown’ in the long term. Examples for
fiscal policy of measuring such resilience
include fiscal space and fiscal stress tests,
both of which are described in more
detail later.
As mentioned above, uncertain events
by definition cannot be directly measured;
market operations the Reserve Bank uses
to steer short-term interest rates. But just
communicating policy is not a panacea.
Poor or unclear communication might
work against policy actions, weakening
their effect. Evidence from monetary
economics, for example, emphasises the
role of good, clear communication in
reinforcing the effects of monetary policy
actions.4
So, good, clear communication
of policy in general is important. Can
the same be said of communication of
risks and uncertainties? Since uncertain
events involve the unknowable, these
events cannot be readily communicated.
Hence, we will focus here on risk. The
communication of macroeconomic risks
faces a similar challenge to the one we
saw in making policy in the face of risks:
communication should not add to existing
risks. So how might communicating risks
help to reduce their material impacts if
realised? Is it even possible to be clear
about risks?
Page 42 – Policy Quarterly – Volume 12, Issue 3 – August 2016
There are two main reasons to believe
that risks should be communicated. The
first is to explicitly acknowledge that
government policy is being made in a
risky world. This includes acknowledging
whether or not policymakers are
explicitly accounting for risks in their
decisions, or are simply making decisions
as if things were certain. Being open in
communicating the particular risks in the
face of which policy was made helps the
public better understand policymakers’
objectives and reasoning. The second is
to help justify any future recalibration
of policy. If risks were not initially
communicated, any subsequent policy
adjustments may appear to be an ‘about
face’. This greater public understanding
is likely to make subsequent policy
adjustments easier to implement
politically.
Take earthquake insurance as an
example of the second point. Suppose
the risk of an earthquake occurring is
viewed as having increased, because, say,
of a better scientific understanding of the
causes of seismic activity. This suggests
that the insurance premiums should rise.
If the risk and its subsequent increase
have been communicated, then it will
be understood that this rise has not
involved any change in policy, but simply
reflects the same policy being updated
to reflect evolving circumstances. The
understanding of higher risk of an
earthquake is also likely to trigger a
market response, such as better household
preparedness or changes in the building
code.
But can regular communication help
make things clearer and reduce the level
of risk? One example of communication
reducing the level of risk is from exchange
rate regimes. Fixed exchange rates in one
sense should be more predictable, but
in practice they are subject to periodic
large and sudden revaluations. Allowing
the exchange rate to float – so that it
reflects existing market perceptions of
risk – can actually make its movements
more predictable. It communicates the
necessary (price) information for the
market participants: the private sector
accepts and adapts to the exchange
rate risks as part of its business
environment. Floating the exchange rate
Tools to communicate risks
This section examines some examples
of informal and formal (statistical)
Figure 1: An illustrative fan chart for RBNZ forecast of 90-day interest rate
9
8
7
6
Percent, p.a.
eventually reduces the risk of building
macroeconomic imbalances and systemic
vulnerability to crisis (Ghosh and Ostry,
2009).
A further case for communicating risks
is simply that the analysis of these risks
will often already have been undertaken as
part of policymaking, and this information
may be valuable in and of itself.
Of course, there may be a limit to how
much can and should be communicated.
The United States Federal Reserve board
of governors’ choice to publish their
Tealbooks and Bluebooks with a time lag
is an example of constraining the degree
of communication in order to protect the
quality of policy deliberations. A former
vice-chairman, Don Kohn, has argued
that although prompt publication of such
documents may be useful from the public’s
perspective, it is not so clear that it is
desirable from the institution’s perspective.
The main concern was that the board staff
would be more cautious (and thus less
open) in putting their recommendations
forward if they knew they were going to
be made public with the decision.
We now turn to economic resilience.
We feel that evaluation and communication of resilience, both qualitative and
quantitative, would be beneficial for two
reasons. First, the need to communicate
issues of resilience to the public would
help focus policymakers on thinking
through the issue clearly. Second, and
most importantly given the difficulties
relating to resilience to uncertainty, it
will allow for open feedback from the
general public. Some of this feedback will
be in the form of direct submissions to
government agencies, but much is likely
to be simply public discussion in the news
media and online. Discussion of resilience
is likely to be mostly qualitative, given
the unquantifiable nature of uncertainty.
There will be a lesser role for quantitative
assessments, such as measuring fiscal space
and implementing stress tests. Stress tests
involve quantitatively looking at how the
economy would react to combinations of
major shocks which would be large by
historical standards.
5
4
3
2
1
0
2012
2013
2014
2015
2016
Quarters
RBNZ’s forecast, MPS March 2016
Probabilistically weighted alternative scenarios
2017
2018
2019
History
90% confidence interval (5ppt increments)
Source: Reserve Bank of New Zealand; authors’ calculations
Note: The fan chart in this figure is for illustrative purposes only. The Reserve Bank does not publish its forecast with any measure of
confidence. We construct the confidence intervals using a vector autoregressive model of CPI quarterly inflation, GDP quarterly
growth, 90-day rate and quarterly percentage change in the real trade-weighted index of exchange rates. The model was
estimated on quarterly New Zealand data from 1995q1 through 2015q4.
communication of risks in the practice
of monetary and fiscal policy in New
Zealand, the United Kingdom and the
US; we also draw some general lessons for
policy more broadly.
Verbal communication
Central banks have been at the forefront in
communicating policy. For instance, the
Reserve Bank of New Zealand publishes a
monetary policy statement accompanying
every decision on setting interest rates,
explaining the bank’s views on the state
of the economy and why it made the
decision it did.5 These statements include
some discussion of both uncertainty
and risk, but do not extend to a formal
evaluation of risks. There is also typically
an accompanying series of research papers
and notes providing deeper analyses of
specific topics.
Other central banks take this a
step further and release the minutes of
their policy deliberations (for example,
the Bank of England, the US Federal
Reserve System and the Reserve Bank
of Australia). Releasing the minutes is
seen not only as a way to communicate
why certain policy decisions were made;
by revealing internal disagreements at
these meetings the minutes also provide
a gauge as to how much confidence
there was around reaching the final
decision and guide expectations about
risks surrounding future economic
outcomes. Some market analysts set
up word-counters: if statements are
longer than average, this may signal that
decisions were hard to make; repetition
of particular words may signal specific
policy biases. This informal measure
of risk is seen as an important part of
communicating the monetary policy
decisions of these central banks.
We turn now to considering a few
examples of more formal analysis (with
an acknowledgement that these methods
are appropriate only for risk and not
for uncertainty): the use of fan charts,
alternative scenarios and identification of
the nature and sources of risks.
Fan charts
One example of formally communicating risk is the Bank of England’s fan
charts.6 Each quarter the Bank of
England releases an inflation report7
which contains forecasts of short-term
interest rates, inflation or GDP growth,
forecasts that are not just a single path
(or ‘baseline forecast’). ‘Uncertainty’
around the forecasts is shown in the
form of fan charts showing a range of
possible future outcomes, along with
Policy Quarterly – Volume 12, Issue 3 – August 2016 – Page 43
Managing and Communicating Risk and Uncertainty in Macroeconomic Policymaking
Table 1: Sources of risks for net core Crown debt projections: illustration
Fiscal year
Macroeconomic risks
Fiscal risks
Statistical error risks
2015
0%
0%
100%
2016
20%
49%
32%
2017
25%
63%
12%
2018
27%
64%
9%
2019
27%
65%
8%
2020
27%
65%
8%
2021
27%
66%
7%
2022
26%
69%
5%
2023
26%
72%
3%
2024
25%
74%
1%
Source: Authors’ calculations
Note: For illustration purposes only. Units are the percentage shares of total risk. The results above are taken from the authors’ work on
measuring uncertainty around the government’s target of reducing the value of net public debt to below 20% of GDP by around
2020.
estimations of how likely they are to
occur.
In Figure 1 we present an illustration
of what a fan chart for the Reserve
Bank of New Zealand’s forecast of the
interest rate on the 90-day bank bill,
the key money market rate, might look
like. The Treasury in its budget and
half-year economic updates regularly
publishes a similar measure of risk for
its public revenue projections. The lefthand side of the figure simply shows
historical outcomes, and so we just have
the solid line. At present all the Reserve
Bank publishes is a point forecast for
interest rates, which is here shown as
the dotted line extending into the future.
But the future is uncertain, and while the
present forecasts simply present a single
future, a fan chart provides much richer
information about other likely future
outcomes. These likely alternative futures
are represented by the shaded regions.
The darker regions represent the central
outcome, with lighter regions indicating
progressively less likely outcomes.
While the Treasury and Reserve
Bank discuss risks in their economic
forecasts, their probabilistic evaluation
is not as systematic as it could be
and it is often limited to a handful
of forecasted variables. We believe
that fan charts, or some other way of
illustrating the probabilistic measure of
confidence about the economic outlook,
should be routinely reported for every
macroeconomic variable that the Treasury
and Reserve Bank forecast. Enhancing
their communication in this way would
benefit public discussion about economic
policy and help build its credibility. Such
measures of confidence will shed light
on how optimistic or pessimistic those
forecasts are.
Alternative scenarios
Another
common
communication
tool is the assessment of plausible
alternative scenarios. Fan charts provide
information about the likely distribution
of economic outcomes for all possible
risks and combinations of them that the
economy faced in the past. However, the
economy might be subject to very specific
headwinds or tailwinds. The alternative
scenarios evaluate the economic impact
of these specific risks. When assigned
probabilistic weights and aggregated, the
alternative scenarios measure helps to
reduce the balance of prevailing risks. We
should stress that the alternative scenarios
are a measure of perceived risks and are
therefore inherently subjective.
At its full width the shaded area in the
fan chart in Figure 1 displays the likely
outcomes in 90% of possible futures, an
assessment based on all the economic
outcomes seen over the past 20 years of
a stable monetary policy regime. The
dashed line represents the probabilistically
weighted alternative economic scenarios.
The purpose of the alternative scenarios is
to help narrow, and point to, skewness in
the uncertainty distribution and inform
the decision-makers about the likely
direction of the risks. In the case shown
Page 44 – Policy Quarterly – Volume 12, Issue 3 – August 2016
in Figure 1, the probabilistic summary of
alternative scenarios points to risks being
skewed to the downside over the forecast
horizon, informing the policymaker that
they are more likely to face downside
than upside risks.
Risk classification and identification
Not all risks can be controlled for or are
worth responding to. For macroeconomic
risk management in practice, we also
need to understand the sources of risk. A
useful tool is a structural macroeconomic
model that allows us to identify the main
structural drivers of risk that underpin fan
charts such as the one above. Using such
models for structural decompositions of
historical risks helps identify risks that
can be fully or partially controlled, and
those that are simply uncontrollable and
must be accepted.
For example, current research on
the sources of risks for the government
net debt projections classifies risks into
three main categories: macroeconomic
uncertainty, statistical uncertainty and
fiscal uncertainty. Table 1 illustrates
the risk classification. Governments
have no control over statistical risk (for
example, GDP figures are subject to
data revisions). Governments have only
modest control over macroeconomic
risk arising from external sources, such
as the level of Chinese demand, the
exchange rate, foreign competition,
droughts and so on. The level of fiscal
policy risks is, however, under the
government’s control, especially over
longer periods of time.
The risk classification yields useful
insights. For example, most of the risks
related to statistical confidence are
over short time horizons, and thus the
government could not control or reduce
these risks. Policymakers may, however,
choose to create a buffer capable of
absorbing the consequences of these
risks, and the risk identification and
quantification allows us to evaluate the size
such a buffer would need to be. Or they
can communicate risks by making clear
that debt level targets are best understood
as a general goal, not as something to be
hit with precision. By contrast, over longer
time horizons the majority of the risks
fall into the category of fiscal policy risks
(i.e. risks around tax revenue outcomes or
expenditure policy), and are thus within
the government’s ability to control and
reduce by communication or guidance.
Conclusion
We conclude that greater awareness
of the risk landscape in which policy
decisions are made leads to better policy,
and allows for the possibility of better
communication of policy decisions to
the public. The Treasury and Reserve
Bank of New Zealand currently include
a discussion of various risks in their
public communications, but would
benefit from better and more systematic
communication of these risks to the
public. Publishing measures of statistical
confidence for economic projections
or assumptions underlying economic
policies should be the new practice. Doing
so will require an effort to separate what
events are considered to be quantifiable
risks and what are uncertainties.
Adoption of some of the tools for
formally analysing risk that we have
discussed may be difficult for institutions
without advanced analytical capabilities.
But it represents a perfect opportunity
for cooperation between government
agencies and universities. Academic
research can help to advance better
government in this area by providing key
analytical capability and build internal
government capability through education
and training.
1 There is also an aspect of ambiguity in the context of
risks and uncertainties that can give rise to heterogeneous
expectations (Hansen and Sargent, 2012), but we will leave
this aspect out of the present discussion.
2 For example, consider the ‘debt ceilings’ in the United
States. These impose a dollar limit on the amount of debt
that the US government can issue. But Congress can
approve a budget for spending and revenue that would
necessitate an amount of borrowing in excess of this limit.
This creates a risk relating to what would actually happen if
the debt ceiling is not raised, and this risk is created by the
policies themselves.
3 This latter point of reducing fluctuations in GDP should
not, however, be over-interpreted as saying that the level
of uncertainty is more than a minor contributing factor to
recessions (Bachmann and Bayer, 2013; Bachmann, Elstner
and Sims, 2013; Jurado, Ludvigson and Ng, 2015).
4 See http://www.voxeu.org/article/central-bankcommunication; http://www.voxeu.org/article/measuringclarity-central-bank-communication.
5 The Reserve Bank has been releasing monetary policy
statements going back to 1996: see http://www.rbnz.govt.
nz/monetary_policy/monetary_policy_statement/.
6 The Bank of England was one of the first central banks to
introduce this communication tool.
7 http://www.bankofengland.co.uk/publications/Pages/
inflationreport/default.aspx.
References
Bachmann, R. and C. Bayer (2013) ‘“Wait-and-see” business cycles?’,
Journal of Monetary Economics, 60 (6), pp.704-19
Bachmann, R., S. Elstner and E.R. Sims (2013) ‘Uncertainty and
economic activity: evidence from business survey data’, American
Economic Journal: Macroeconomics, 5 (2), pp.217-49
Bekaert, G., M. Hoerova and M.L. Duca (2013) ‘Risk, uncertainty and
monetary policy’, Journal of Monetary Economics, 60 (7), pp.771-88
Bertola, G., L. Guiso and L. Pistaferri (2005) ‘Uncertainty and consumer
durables adjustment’, Review of Economic Studies, 72 (4), pp.9731007
Bloom, N. (2009) ‘The impact of uncertainty shocks’, Econometrica, 77
(3), pp.623-85
Bloom, N., S. Bond and J. Van Reenen (2007) ‘Uncertainty and
investment dynamics’, Review of Economic Studies, 74 (2), pp.391415
Dupuis, P., M.R. James and I.R. Peterson (2000) ‘Minimax optimal
control of stochastic uncertain systems with relative entropy
constraints’, IEEE Transactions on Automatic Control, 45( 3), pp.398412
Ilut, C.L. and M. Schneider (2014) ‘Ambiguous business cycles’,
American Economic Review, 104 (8), pp.2368-99
International Monetary Fund (2016) Analyzing and Managing Fiscal Risks,
policy paper, Washington: IMF
Jurado, K., S.C. Ludvigson and S. Ng (2015) ‘Measuring uncertainty’,
American Economic Review, 105 (3), pp.1177-216
Knight, F.H. (1921) Risk, Uncertainty, and Profit, Boston: Hart, Schaffner
& Marx and Houghton Mifflin
Kydland, F. and E.C. Prescott (1977) ‘Rules rather than discretion: the
inconsistency of optimal plans’, Journal of Political Economy, 85 (3)
pp.473-92
Manski, C. (2013) Public Policy in an Uncertain World: analysis and
decisions, Cambridge, Mass; London: Harvard University Press
Manski, C. (2015) ‘Communicating uncertainty in official economic
statistics: an appraisal fifty years after Morgenstern’, Journal of
Economic Literature, 53 (3), pp.631-53
Orphanides, A. (2001) ‘Monetary policy rules based on real-time data’,
American Economic Review, 91 (4), pp.964–85
Svensson, L.E.O. (2009) ‘What have economists learned about monetary
Fernández-Villaverde, J., P. Guerrón-Quintana, K. Kuester and J. Rubio-
policy over the last fifty years’, in Deutsche Bundesbank (ed.),
Ramírez (2015) ‘Fiscal volatility shocks and economic activity’,
Monetary Policy Over Fifty Years: experience and lessons, Oxon:
American Economic Review, 105 (11), pp.3352-84
Routledge
Ghosh, A.R. and J.D. Ostry (2009) ‘Choosing an exchange rate regime’,
Finance and Development, 46 (4), pp.38-40
Hansen, L.P. and T.J. Sargent (2001) ‘Robust control and model
uncertainty’, American Economic Review, 91 (2), pp.60-6
Ter-Minassian, T. (2014) External Review of the Treasury’s Fiscal
Policy Advice: New Zealand, report prepared for the New Zealand
Treasury, available at http://www.treasury.govt.nz/publications/
informationreleases/fiscalpolicyadvice/pdfs/tfpa-2908566.pdf
Hansen, L.P. and T.J. Sargent (2012) ‘Three types of ambiguity’, Journal
of Monetary Economics, 59 (5), pp.422-45
Policy Quarterly – Volume 12, Issue 3 – August 2016 – Page 45